Related Experiment Videos
Quantifying the bias associated with use of discrepant analysis
1Division of Laboratory Systems, Public Health Practice Program Office, Centers for Disease Control and Prevention, Atlanta, GA 30341-3714, USA. hx10@cdc.gov
Clinical Chemistry
|April 29, 1998
Summary
Discrepant analysis for laboratory tests often overestimates performance due to bias. This bias is influenced by test characteristics and specimen proportions, suggesting alternative methods may be needed.
Area of Science:
- Clinical chemistry
- Laboratory medicine
- Diagnostic test evaluation
Background:
- Discrepant analysis is a common method for evaluating laboratory test performance.
- It involves comparing a candidate test against a comparison method, with discrepancies resolved by a confirmatory test.
Purpose of the Study:
- To quantify the bias inherent in discrepant analysis.
- To identify factors influencing this bias in laboratory test performance estimation.
Main Methods:
- Quantification of bias in discrepant analysis.
- Analysis of the influence of analytical sensitivity and specificity of candidate, comparison, and confirmatory tests.
- Assessment of the impact of the proportion of abnormal specimens.
Main Results:
- Discrepant analysis typically yields biased estimates, often positive, leading to overestimation of test performance parameters.
- The bias magnitude and direction are predictable based on the analytical sensitivity and specificity of the involved tests.
- The proportion of abnormal specimens significantly impacts bias, especially for analytical sensitivity and positive predictive value estimates.
Conclusions:
- Discrepant analysis commonly overestimates laboratory test performance parameters.
- Understanding the factors influencing bias is crucial for accurate test evaluation.
- Alternative analytical approaches may be necessary to mitigate bias in discrepant analysis.